Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.8.1
Description
The performance_schema counter wait/io/file/innodb/innodb_log_file is always reported as 0.
The way how redo log writes are being waited for was refactored in MDEV-21534, which is when the uninstrumented flush_lock and write_lock were introduced. Even before this change, all the wait/io/file/innodb/ counters were always 0 in my tests. I tested the latest 10.2 yesterday.
Moreover, if the PMEM interface that was introduced in MDEV-17084 is being used, writes to the InnoDB log file will completely avoid any system calls and performance_schema instrumentation. The PMEM interface was rewritten in MDEV-14425 to cover also reads (during recovery).
In MDEV-26603 we plan to implement asynchronous writes for the redo log.
When it comes to persistent data files (wait/io/file/innodb/innodb_data_file) and the temporary data file (wait/io/file/innodb/innodb_temp_file), all writes are already since MDEV-23855 asynchronous and thus uninstrumented by performance_schema. Only page reads on a buffer pool miss would be synchronous and thus theoretically coverable by performance_schema. But, this ticket is only about removing the ib_logfile0 redo log file instrumentation.
Attachments
Issue Links
- relates to
-
MDEV-14425 Change the InnoDB redo log format to reduce write amplification
- Closed
-
MDEV-17084 Optimize append only files for NVDIMM
- Closed
-
MDEV-21534 improve locking/waiting in log_write_up_to
- Closed
-
MDEV-26603 asynchronous redo log write
- Closed